Symmetric task

Reading libraries and parameters

library(tidyverse)
library(quickpsy)
library(cowplot)
list.files("R", full.names = TRUE) %>% walk(source)
source("graphical_parameters.R")
source("parameters.R")
load(file = "logdata/dat_asym.RData")

Fixed lapse rate to 0% or 1%

fun_normal<- function(x, p) {
  pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
}
fun_normal_asyn_one <- function(x, p) {
  .01 + (1- .01) * pnorm(x, p[1] - p[2], p[3]) - (1- .01) * pnorm(x, p[1] + p[2], p[3])
}
fit_asym_without_asyn_zero <- quickpsy(dat_asym, 
                     orientation, response, 
                     fun = fun_normal,
                     parini = list(pini_origin_asym, pini_origin_asym, pini_scale),
                     grouping = .(subject, reference),
                     bootstrap = "none")
fit_asym_without_asyn_one <- quickpsy(dat_asym, 
                             orientation, response, 
                             fun = fun_normal_asyn_one,
                             parini = list(pini_origin_asym, pini_origin_asym, pini_scale),
                             grouping = .(subject, reference),
                             bootstrap = "none")
ms_asym_without_asyn_zero_vs_without_asyn_one <- model_selection_aic(
  fit_asym_without_asyn_zero$aic, fit_asym_without_asyn_one$aic) %>% 
  mutate(best = if_else(best == "first", 
                        "without_asyn_zero", "without_asyn_one"))
ms_asym_without_asyn_zero_vs_without_asyn_one %>% 
  group_by(best) %>% 
  count() 

ggplot() + facet_grid(subject ~ reference) +
  geom_point(data = fit_asym_without_asyn_zero$averages, 
             aes(x = orientation, y = prob)) +
  geom_line(data = fit_asym_without_asyn_zero$curves, 
            aes(x = x, y = y, color = "without asyn zero")) +
  geom_line(data = fit_asym_without_asyn_one$curves, 
            aes(x = x, y = y, color = "without asyn one")) +
  geom_text(data = ms_asym_without_asyn_zero_vs_without_asyn_one,
            aes(x = .7, y = .1, label = best), size = 3) +
  theme_grey() + theme(legend.position = "top") 

Adding a fixed lapse rate does not change much the fits.

Checking that the parameters do not reach the boundaries of the initial parameters

fit_asym_without_asyn_zero$par %>% 
  filter(parn == "p1") %>% 
  filter(abs(par) >= pini_origin_asym[2])

fit_asym_without_asyn_zero$par %>% 
  filter(parn == "p2") %>% 
  filter(abs(par) >= pini_origin_asym[2])

fit_asym_without_asyn_zero$par %>% 
  filter(parn == "p3") %>% 
  filter(abs(par) >= pini_scale[2])

One variable lapse rate vs fixed lapse rate

fun_normal_with_one_asyn <- function(x, p) {
  p[4] + (1 - p[4]) * pnorm(x, p[1] - p[2], p[3]) - (1 - p[4]) * pnorm(x, p[1] + p[2], p[3])
}
fit_asym_with_one_asyn <- quickpsy(dat_asym, 
          orientation, response, 
          fun = fun_normal_with_one_asyn,
          parini = list(pini_origin_asym, pini_origin_asym, pini_scale, pini_lapse),
          grouping = .(subject, reference),
          bootstrap = "none")
fit_asym_without_asyn_zero_better <- ms_asym_without_asyn_zero_vs_without_asyn_one %>% 
  filter(best == "without_asyn_zero") %>% 
  select(subject, reference)
fit_asym_without_asyn_one_better <- ms_asym_without_asyn_zero_vs_without_asyn_one %>% 
  filter(best == "without_asyn_one") %>% 
  select(subject, reference)
                                   
fit_asym_without_asyn_zero_better_logliks <- fit_asym_without_asyn_zero$logliks %>%
  semi_join(fit_asym_without_asyn_zero_better, by = c("subject", "reference"))
  
fit_asym_without_asyn_one_better_logliks <- fit_asym_without_asyn_one$logliks %>%
  semi_join(fit_asym_without_asyn_one_better, by = c("subject", "reference"))
fit_asym_without_asyn_logliks <- fit_asym_without_asyn_zero_better_logliks %>% 
  bind_rows(fit_asym_without_asyn_one_better_logliks)
  
ms_asym_with_one_asyn_vs_without_asyn <- model_selection_lrt(
  fit_asym_with_one_asyn$logliks, fit_asym_without_asyn_logliks)  %>% 
  mutate(best = if_else(best == "first", 
                        "with_one_asyn", "without_asyn"))
fit_asym_with_one_asyn_better <- ms_asym_with_one_asyn_vs_without_asyn %>% 
  filter(best == "with_one_asyn") %>% 
  select(subject, reference)
fit_asym_without_asyn_zero_better_curves <- fit_asym_without_asyn_zero$curves %>%
  semi_join(fit_asym_without_asyn_zero_better, by = c("subject", "reference"))
  
fit_asym_without_asyn_one_better_curves <- fit_asym_without_asyn_one$curves %>%
  semi_join(fit_asym_without_asyn_one_better, by = c("subject", "reference"))
fit_asym_without_asyn_curves <- fit_asym_without_asyn_zero_better_curves %>% 
  bind_rows(fit_asym_without_asyn_one_better_curves)
ggplot() + facet_grid(subject ~ reference) +
  geom_point(data = fit_asym_with_one_asyn$averages, 
             aes(x = orientation, y = prob)) +
  geom_line(data = fit_asym_with_one_asyn$curves, 
            aes(x = x, y = y, color = "one asyn")) +
  geom_line(data = fit_asym_without_asyn_curves, 
            aes(x = x, y = y, color = "without asyn")) +
  geom_text(data = ms_asym_with_one_asyn_vs_without_asyn, 
            aes(x = .7, y = .1, label = best), size = 3) +
  theme_grey() + theme(legend.position = "top") 

Adding a lapse rate parameter produce poor fits.

Fitting

refs <- dat_asym %>% distinct(vertical, references, reference)
fit_asym_full_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  fun_2 <- function(x, p) pnorm(x, p[1] - p[4], p[6]) - pnorm(x, p[1] + p[5], p[6])
  
  pini <- list(pini_origin_asym, pini_origin_asym, pini_scale, 
               pini_origin_asym, pini_origin_asym, pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_same_slope_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  fun_2 <- function(x, p) pnorm(x, p[1] - p[4], p[3]) - pnorm(x, p[1] + p[5], p[3])
  
  pini <- list(pini_origin_asym, pini_origin_asym, pini_scale, 
               pini_origin_asym, pini_origin_asym)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_sym_crit_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  fun_2 <- function(x, p) pnorm(x, p[1] - p[4], p[5]) - pnorm(x, p[1] + p[4], p[5])
  
  pini <- list(pini_origin_asym, pini_origin_asym, pini_scale, 
               pini_origin_asym, pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_sym_crit_same_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  fun_2 <- function(x, p) pnorm(x, p[1] - p[2], p[4]) - pnorm(x, p[1] + p[2], p[4])
  
  pini <- list(pini_origin_asym, pini_origin_asym, pini_scale, 
               pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_sym_crit_same_zero_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, - p[1], p[2]) - pnorm(x, p[1], p[2])
  fun_2 <- function(x, p) pnorm(x, - p[1], p[3]) - pnorm(x, p[1], p[3])
  
  pini <- list(pini_origin_asym, pini_scale, pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_same_slope_sym_crit_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  fun_2 <- function(x, p) pnorm(x, p[1] - p[4], p[3]) - pnorm(x, p[1] + p[4], p[3])
  
  pini <- list(pini_origin_asym, pini_origin_asym, pini_scale, 
               pini_origin_asym)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_same_slope_sym_crit_zero_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, - p[1], p[2]) - pnorm(x,p[1], p[2])
  fun_2 <- function(x, p) pnorm(x, - p[3], p[2]) - pnorm(x, p[3], p[2])
  
  pini <- list(pini_origin_asym, pini_scale, 
               pini_origin_asym)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_same_slope_sym_crit_same_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  fun_2 <- function(x, p) pnorm(x, p[1] - p[2], p[3]) - pnorm(x, p[1] + p[2], p[3])
  
  pini <- list(pini_origin_asym, pini_origin_asym, pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_same_slope_sym_crit_same_zero_fun <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, - p[1], p[2]) - pnorm(x, p[1], p[2])
  fun_2 <- function(x, p) pnorm(x, - p[1], p[2]) - pnorm(x, p[1], p[2])
  
  pini <- list(pini_origin_asym, pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_full_fun_zero <- function(data) {
  
  fun_1 <- function(x, p) pnorm(x, - p[1], p[2]) - pnorm(x, p[1], p[2])
  fun_2 <- function(x, p) pnorm(x, - p[3], p[4]) - pnorm(x, p[3], p[4])
  
  pini <- list(pini_origin_asym, pini_scale, 
               pini_origin_asym, pini_scale)
  fun_df <- data %>% 
    distinct(references) %>% 
    arrange(references) %>% 
    bind_cols(tibble(fun = c(fun_1, fun_2))) 
  quickpsy(data, orientation, response, fun = fun_df, xmin = -4, xmax = 4,
           parini = pini, bootstrap = "none", grouping = .(references),
           threshold = FALSE)
}
fit_asym_best <- dat_asym %>% 
  group_by(subject, vertical) %>% 
  nest() %>% 
  mutate(fit_asym_full = map(data, fit_asym_full_fun),
         fit_asym_full_sym_crit = map(data, fit_asym_full_sym_crit_fun),
         fit_asym_full_sym_crit_same = map(data, fit_asym_full_sym_crit_same_fun),
         fit_asym_full_sym_crit_same_zero = map(data, fit_asym_full_sym_crit_same_zero_fun),
         fit_asym_full_zero = map(data, fit_asym_full_fun_zero),
         fit_asym_full_same_slope = map(data, fit_asym_full_same_slope_fun),
         fit_asym_full_same_slope_sym_crit = map(data, fit_asym_full_same_slope_sym_crit_fun),
         fit_asym_full_same_slope_sym_crit_zero = map(data, fit_asym_full_same_slope_sym_crit_zero_fun),
         fit_asym_full_same_slope_sym_crit_same = map(data, fit_asym_full_same_slope_sym_crit_same_fun),
         fit_asym_full_same_slope_sym_crit_same_zero = map(data, fit_asym_full_same_slope_sym_crit_same_zero_fun)) %>% 
      #   fit_asym_zero = map(data, fit_asym_zero_fun),
       #  fit_asym_p = map(data, fit_asym_p_fun),
       #  fit_asym_d = map(data, fit_asym_d_fun)) %>% 
  select(-data)

Full

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full same slope

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full vs full same slope

asym_full_vs_full_same_slope <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_loglik = map(fit_asym_full, "logliks"),
         fit_asym_full_same_slope_loglik = map(fit_asym_full_same_slope, "logliks"),
         best = map2_chr(fit_asym_full_loglik, fit_asym_full_same_slope_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_vs_full_same_slope %>% 
  group_by(best) %>% 
  count()     

         
best_asym_full_no_same_slope <- asym_full_vs_full_same_slope %>% 
  filter(best == "first") %>% 
  select(subject, vertical)
best_asym_full_same_slope <- asym_full_vs_full_same_slope %>% 
  filter(best == "second") %>% 
  select(subject, vertical)

Full no same slope sym crit

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_sym_crit, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_sym_crit, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full no same slope vs full no same slope sym crit

asym_full_no_same_slope_vs_full_no_same_slope_sym_crit <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_loglik = map(fit_asym_full, "logliks"),
         fit_asym_full_sym_crit_loglik = map(fit_asym_full_sym_crit, "logliks"),
         best = map2_chr(fit_asym_full_loglik, fit_asym_full_sym_crit_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_no_same_slope_vs_full_no_same_slope_sym_crit %>% 
  semi_join(best_asym_full_no_same_slope) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
best_asym_full_no_same_slope_sym_crit <- asym_full_no_same_slope_vs_full_no_same_slope_sym_crit %>% 
  semi_join(best_asym_full_no_same_slope) %>% 
  filter(best == "second") %>% 
  select(subject, vertical)

Joining, by = c("subject", "vertical")

Full no same slope sym crit same

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_sym_crit_same, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_sym_crit_same, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full no same slope sym crit vs full no same slope sym crit same

asym_full_no_same_slope_sym_crit_vs_full_no_same_slope_sym_crit_same <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_sym_crit_loglik = map(fit_asym_full_sym_crit, "logliks"),
         fit_asym_full_sym_crit_same_loglik = map(fit_asym_full_sym_crit_same, "logliks"),
         best = map2_chr(fit_asym_full_sym_crit_loglik, fit_asym_full_sym_crit_same_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_no_same_slope_sym_crit_vs_full_no_same_slope_sym_crit_same %>% 
  semi_join(best_asym_full_no_same_slope_sym_crit) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
best_asym_full_no_same_slope_sym_crit_same <- asym_full_no_same_slope_sym_crit_vs_full_no_same_slope_sym_crit_same %>% 
  semi_join(best_asym_full_no_same_slope_sym_crit) %>% 
  filter(best == "second") %>% 
  select(subject, vertical)

Joining, by = c("subject", "vertical")

Full no same slope sym crit same zero

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_sym_crit_same_zero, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_sym_crit_same_zero, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full no same slope sym crit same vs full no same slope sym crit same zero

asym_full_no_same_slope_sym_crit_same_vs_full_no_same_slope_sym_crit_same_zero <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_sym_crit_same_loglik = map(fit_asym_full_sym_crit_same, "logliks"),
         fit_asym_full_sym_crit_same_zero_loglik = map(fit_asym_full_sym_crit_same_zero, "logliks"),
         best = map2_chr(fit_asym_full_sym_crit_same_loglik, fit_asym_full_sym_crit_same_zero_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_no_same_slope_sym_crit_same_vs_full_no_same_slope_sym_crit_same_zero %>% 
  semi_join(best_asym_full_no_same_slope_sym_crit_same) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
### Add to s vs d
best_asym_full_no_same_slope_sym_crit_same_no_zero <- asym_full_no_same_slope_sym_crit_same_vs_full_no_same_slope_sym_crit_same_zero %>%
  semi_join(best_asym_full_no_same_slope_sym_crit_same) %>% 
  filter(best == "first") %>% 
  select(subject, vertical) %>% 
  mutate(best = "response")

Joining, by = c("subject", "vertical")

Full same slope sym crit

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full same slope vs full same slope sym crit

asym_full_same_slope_vs_full_same_slope_sym_crit <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_same_slope_loglik = map(fit_asym_full_same_slope, "logliks"),
         fit_asym_full_same_slope_sym_crit_loglik = map(fit_asym_full_same_slope_sym_crit, "logliks"),
         best = map2_chr(fit_asym_full_same_slope_loglik, fit_asym_full_same_slope_sym_crit_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_same_slope_vs_full_same_slope_sym_crit %>% 
  semi_join(best_asym_full_same_slope) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
### Add to s vs d
best_asym_full_same_slope_no_sym_crit <- asym_full_same_slope_vs_full_same_slope_sym_crit %>%
  semi_join(best_asym_full_same_slope) %>% 
  filter(best == "first") %>% 
  select(subject, vertical) %>% 
  mutate(best = "no sym crit")

Joining, by = c("subject", "vertical")

best_asym_full_same_slope_sym_crit <- asym_full_same_slope_vs_full_same_slope_sym_crit %>%
  semi_join(best_asym_full_same_slope) %>% 
  filter(best == "second") %>% 
  select(subject, vertical) 

Joining, by = c("subject", "vertical")

Full same slope sym crit same

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit_same, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit_same, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full same slope sym crit vs full same slope sym crit same

asym_full_same_slope_sym_crit_vs_full_same_slope_sym_crit_same <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_same_slope_sym_crit_loglik = map(fit_asym_full_same_slope_sym_crit, "logliks"),
         fit_asym_full_same_slope_sym_crit_same_loglik = map(fit_asym_full_same_slope_sym_crit_same, "logliks"),
         best = map2_chr(fit_asym_full_same_slope_sym_crit_loglik, fit_asym_full_same_slope_sym_crit_same_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_same_slope_sym_crit_vs_full_same_slope_sym_crit_same %>% 
  semi_join(best_asym_full_same_slope_sym_crit) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
best_asym_full_same_slope_sym_crit_no_same <- asym_full_same_slope_sym_crit_vs_full_same_slope_sym_crit_same %>%
  semi_join(best_asym_full_same_slope_sym_crit) %>% 
  filter(best == "first") %>% 
  select(subject, vertical) 

Joining, by = c("subject", "vertical")

best_asym_full_same_slope_sym_crit_same <- asym_full_same_slope_sym_crit_vs_full_same_slope_sym_crit_same %>%
  semi_join(best_asym_full_same_slope_sym_crit) %>% 
  filter(best == "second") %>% 
  select(subject, vertical) 

Joining, by = c("subject", "vertical")

Full same slope sym crit no same zero

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit_zero, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit_zero, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full same slope sym crit no same vs full same slope sym crit no same zero

asym_full_same_slope_sym_crit_no_same_vs_full_same_slope_sym_crit_no_same_zero <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_same_slope_sym_crit_loglik = map(fit_asym_full_same_slope_sym_crit, "logliks"),
         fit_asym_full_same_slope_sym_crit_zero_loglik = map(fit_asym_full_same_slope_sym_crit_zero, "logliks"),
         best = map2_chr(fit_asym_full_same_slope_sym_crit_loglik, fit_asym_full_same_slope_sym_crit_zero_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_same_slope_sym_crit_no_same_vs_full_same_slope_sym_crit_no_same_zero %>% 
  semi_join(best_asym_full_same_slope_sym_crit_no_same) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
### Add to s vs d
best_asym_full_same_slope_sym_crit_no_same_no_zero <- asym_full_same_slope_sym_crit_no_same_vs_full_same_slope_sym_crit_no_same_zero %>%
  semi_join(best_asym_full_same_slope_sym_crit_no_same) %>% 
  filter(best == "first") %>% 
  select(subject, vertical) %>% 
  mutate(best = "response")

Joining, by = c("subject", "vertical")

Full same slope sym crit same zero

ggplot() + facet_wrap(subject ~ vertical, ncol = 6) +
  geom_point(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit_same_zero, "averages")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
             aes(x = orientation, y = prob, color = references)) +
  geom_line(data = fit_asym_best %>% 
               mutate(temp = map(fit_asym_full_same_slope_sym_crit_same_zero, "curves")) %>% 
               select(subject, vertical, temp) %>% 
               unnest(temp), 
            aes(x = x, y = y, color = references)) +
  theme_grey() + theme(legend.position = "top") 

Full same slope sym crit same vs full same slope sym crit same zero

asym_full_same_slope_sym_crit_same_vs_full_same_slope_sym_crit_same_zero <- fit_asym_best %>% 
  group_by(subject, vertical) %>% 
  transmute(fit_asym_full_same_slope_sym_crit_same_loglik = map(fit_asym_full_same_slope_sym_crit_same, "logliks"),
         fit_asym_full_same_slope_sym_crit_same_zero_loglik = map(fit_asym_full_same_slope_sym_crit_same_zero, "logliks"),
         best = map2_chr(fit_asym_full_same_slope_sym_crit_same_loglik, fit_asym_full_same_slope_sym_crit_same_zero_loglik, 
                    ~model_selection_lrt(.x, .y)$best))
asym_full_same_slope_sym_crit_same_vs_full_same_slope_sym_crit_same_zero %>% 
  semi_join(best_asym_full_same_slope_sym_crit_same) %>% 
  group_by(best) %>% 
  count()     

Joining, by = c("subject", "vertical")

         
### Add to s vs d
best_asym_full_same_slope_sym_crit_same_no_zero <- asym_full_same_slope_sym_crit_same_vs_full_same_slope_sym_crit_same_zero %>%
  semi_join(best_asym_full_same_slope_sym_crit_same) %>% 
  filter(best == "first") %>% 
  select(subject, vertical) %>% 
  mutate(best = "sensory")

Joining, by = c("subject", "vertical")

### Add to s vs d
best_asym_full_same_slope_sym_crit_same_zero <- asym_full_same_slope_sym_crit_same_vs_full_same_slope_sym_crit_same_zero %>%
  semi_join(best_asym_full_same_slope_sym_crit_same) %>% 
  filter(best == "second") %>% 
  select(subject, vertical) %>% 
  mutate(best = "zero")

Joining, by = c("subject", "vertical")

Averages, curves and parameters

Add all best

best_asym <- best_asym_full_no_same_slope_sym_crit_same_no_zero %>% 
  bind_rows(best_asym_full_same_slope_no_sym_crit) %>% 
  bind_rows(best_asym_full_same_slope_sym_crit_no_same_no_zero) %>% 
  bind_rows(best_asym_full_same_slope_sym_crit_same_no_zero) %>%  
  bind_rows(best_asym_full_same_slope_sym_crit_same_zero) 
refs <- dat_asym %>% distinct(vertical, references, reference)
asym_averages_s_vs_d_best <-  asym_averages_s_vs_d %>%
  left_join(best_asym) %>% 
  left_join(refs)

Joining, by = c("subject", "vertical")
Joining, by = c("vertical", "references")

  
asym_curves_s_vs_d_best <- asym_curves_s_vs_d %>% 
  left_join(best_asym) %>% 
  left_join(refs)

Joining, by = c("subject", "vertical")
Joining, by = c("vertical", "references")

asym_par_s_vs_d_best <- asym_par_s_vs_d %>% 
  left_join(best_asym) 

Joining, by = c("subject", "vertical")

asym_par_s_vs_d_best_long <- asym_par_s_vs_d_long %>% 
  left_join(best_asym) 

Joining, by = c("subject", "vertical")

asym_par_s_vs_d_best_abs <- asym_par_s_vs_d_best_long %>% 
  select(subject, vertical, psensory, pdecisional, best) %>% 
  gather(parn, par, - subject, - vertical, -best) %>% 
  mutate(parn = if_else(parn == "psensory", 
                             "Sensory\nbias", "Decisional\nbias"),
                abs_par = abs(par))

Save data

save(asym_averages_s_vs_d_best, file = "logdata/asym_averages_s_vs_d_best.RData")
save(asym_curves_s_vs_d_best, file = "logdata/asym_curves_s_vs_d_best.RData")
save(asym_par_s_vs_d_best, file = "logdata/asym_par_s_vs_d_best.RData")
save(asym_par_s_vs_d_best_long, file = "logdata/asym_par_s_vs_d_best_long.RData")
save(asym_par_s_vs_d_best_abs, file = "logdata/asym_par_s_vs_d_best_abs.RData")